Release Notes#

Engula v1.0 Release Notes#

Release window: January 2024 – November 2024
Focus: Compression capabilities and dictionary training, observability and metrics, testing and toolchain, stability fixes.

🚀 Highlights#

The v1.0 release delivers compression capabilities along with the accompanying toolchain: it introduces ZSTD/FSST compression and adaptive dictionaries, improves metrics and statistics output, and rounds out the testing and release pipeline.

✨ New Features#

Compression and dictionaries:

• Added FSST support and split FSST/ZSTD build.
• Added adaptive dictionaries for ZSTD compression.
• Added KV compression for hashtable and t_hash compression support.
• Added "MC DICTINFO" command and "MC OBJECT" command for compression inspection.

Observability and metrics:

• Added the engula info command to print internal statistics.
• Added new metrics and confirmed redis-exporter metrics.
• Added metrics and stats for dict recycle and LIST compression.

Cluster and operations:

• Cluster failover done.
• Added bgsave check.

Testing and tooling:

• Added ginkgo framework and updated ginkgo test cases.
• Added TCL test cases support.
• Added init.sh script for setting up development environment.

🛠 Improvements and Refactoring#

Compression and resource reclamation:

• Refactored mc compress and dict train flow.
• Enhanced dict recycle policy and split reclaim process for HT entries.
• Removed confused compress_stats_tot_try_cnt_ metrics.

Licensing and compliance:

• License will be expired 90 days after the compile time.
• Changed the license expire date from 90 days to 30 days.

Release and versioning:

• Updated engula-server-lib version.
• Docs: update engula version to 1.0.1.
• Added engula-server Dockerfile and refined dockerfile/workflow image names.

🐛 Bug Fixes#

• Fixed invalid memory access in compress_cron_task for hash object recycle.
• Fixed wrong database id for ht element.
• Fixed rdb checksum parsing issue.
• Fixed serverPanic issue.
• Fixed exception in human_readable_to_size() when str_size is '0B'.
• Fixed "access after free" issue when AOF reloading.
• Fixed memory double free in hrandfieldWithCountCommand().
• Fixed misc compile errors on Linux/macOS (including hiredis).

🛠 Infrastructure and CI#

• Added workflow to auto release cross-platform tarball on tag.
• Added weekly auto build.
• Updated GitHub Actions workflows and build docker images.

Engula v2.0 Release Notes#

Release window: December 2024 – March 2025
Focus: Core memory engine integration, basic defragmentation, data structure support.

🚀 Highlights#

This release introduces the eallocator memory engine and integrates it deeply with Valkey, laying a foundation for high-performance memory management: including ZSTD compression, Memory Block defragmentation, and dedicated handling for complex data types (Hash, Set, Listpack).

✨ New Features#

Core engine integration:

• Integrated Valkey with eallocator and wrapped the key-value store into ekvstore.
• Introduced eallocator support for basic memory arena management.
• Added support for Mutable and Immutable memory blocks.

Compression and encoding:

• Added support for ZSTD compression in earena.
• Implemented Variable Byte Encoding for integers.

Data structure support:

• Added eallocator support for Hash objects and Set objects.
• Refactored Listpack to support Set and Hash objects efficiently.
• Implemented eobj serialization and deserialization.

Observability:

• Introduced the engula info command to print internal statistics.

🛠 Improvements and Refactoring#

Defragmentation:

• Rolled out Memory Defragmentation Phase 2 (merging blocks) and Phase 3.
• Added background threads specifically for zip and defragmentation tasks.
• Optimization: Empty blocks are no longer added to the defrag request list.

Memory management:

• Supported up to 32KB record contents in dense/sparse memory blocks.
• Optimized async deletion using a waitable queue in batch mode to reduce CPU usage.
• Refactored elpobj pointers.

🐛 Bug Fixes#

• Fixed ZUNION errors when mixing Set and ZSet types.
• Fixed bugs related to HINCRBY results not saving to listpack.
• Resolved memory leaks in e-record and fixed "access after free" issues.
• Fixed Universal Test (UT) errors for bit operations.

Engula v2.1 Release Notes#

Release window: April 2025 – November 2025
Focus: Performance tuning, large Value support, module system, upstream synchronization (Valkey 8.1 / Redis 7.2).

🚀 Highlights#

The 2.1 release focuses on performance and scalability: it includes significant performance gains such as I/O thread and hash table optimizations, and adds support for Values larger than 4GB. It also aligns compatibility with the latest upstream versions of Valkey (8.1.4) and Redis (7.2).

✨ New Features#

Scalability and storage:

• Added support for values larger than 4GB.
• Implemented Active Defragmentation.
• Added support for compress groups.

Module system:

• Added Engula module support.
• Implemented Module GIL (Global Interpreter Lock) support.

Upstream compatibility:

• Merged Valkey 8.1.4 and Valkey 8.1.1.
• Adapted codebase to support Redis 7.2 and related unit tests.

⚡ Performance Improvements#

I/O and threading:

• Performance improvements for 8 IO threads and 2 IO threads.
• Added async IO notify functionality.

Hash table optimization:

• Improved overall performance for hashtable GET operations.
• Optimized rehash performance and reduced decompression counts for HMGET.
• Implemented memory prefetch for rehash operations.

Memory and latency:

• Reduced p99 latency.
• Implemented quick async free.
• Disabled lazyfree_lazy_user_flush by default to optimize flushing behavior.
• Supported 64KB page size for ARM64 architecture.

🛠 Infrastructure and CI#

Build system:

• Added automated packaging for Linux EL7 (CentOS 7) tarballs.
• Added montplex/ubuntu-build docker image for actions.
• Updated GitHub Actions workflows significantly.

🐛 Bug Fixes#

• Fixed VM_HSet bug in the module system.
• Fixed issues related to test scripts.
• Addressed compilation tasks in eallocator.